一种去中心化的在线边缘任务调度与资源分配方法

A Decentralized Online Edge Task Scheduling and Resource Allocation Approach

创新移动应用迅速发展和5G通讯技术的成熟落地引发了终端用户对计算资源下沉至边缘的巨大需求,从而推动了多接入边缘计算(Multi-access Edge Computing,MEC)概念的提出和发展.在MEC环境中,用户可以将移动任务卸载到周围部署的边缘服务器上来加速移动应用.尽管边缘服务器可以在近用户端提供低时延、高响应性的服务,但其仍面临计算资源有限和用户请求多样带来的挑战,因此需要效率高、实时性强的任务调度与资源分配策略来合理地利用边缘资源.然而,目前针对MEC环境下任务调度和资源分配的方法大多基于中心化架构,并以批处理的方式对某个时间段到达任务进行统一调度与资源分配,因此在面对异构复杂的MEC资源网络和高时延敏感用户需求时具有一定的局限性,此外传统方法还缺少对任务多样性和优先级的考量.针对上述问题,本文提出了一种去中心化的在线任务调度与资源分配方法D-TSRA,该方法以任务优先级加权的卸载响应时间为评价指标,由实时任务调度策略、边缘资源分配策略、和运行时任务迁移策略组成.基于真实边缘环境下数据集的实验表明本文所提出的D-TSRA方法与传统任务调度与资源分配方法相比最多能够减少34.21%的加权任务卸载响应时间.

The rapid development of novel mobile applications and advanced communication technologies have created a huge demand for mobile computing resources,which promotes the emergence of Multi-access Edge Computing(MEC)paradigm.In MEC environments,edge users are allowed to offload their mobile tasks to nearby edge servers to relieve the shortage of mobile resources.Through edge servers could provide low-latency services with high-responsible computing capabilities,they are still faced with the challenge of limited computing resources and the massive offloading requests.Thus,smart task scheduling and resource provision strategies with high real-time property are in high demand for better resource utilization.Traditional approaches are usually based on the centralized architecture and batch-processing scheduling mode,which might suffer from low efficiency and high communication overhead.In this paper,we target at the online edge task scheduling and resource allocation problem,propose a decentralized approach,and prove it is O(1/ε2)-competitive when augmented with ε additional computing power.Experiments based on real-world edge environments have demonstrated that the proposed approach could achieve at most 34.21%reduction on the average weighted offloading response time.